10 Replies Latest reply on Feb 13, 2015 11:39 AM by Jared Conway

    What's important?

    Steve Calkins

      Hi All,

       

      I've been running a model, making adjustments and re-running, trying to get results that match up with my actual test results.

       

      The model is of a glass plate being cooled by air flow across it.  I have the real-life machine in the shop, and have measured the temperature drop in it with a TC-equipped glass plate.  As far as I can tell, the model is as close to reality as I can get it.

       

      However, my real-life results are radically different from the flow model results.  In real-life, my glass plate starts out at about 150C, and is at about 25C after air blowing across it for 2-1/2 minutes.  In the simulation, it only comes down to 80C.

       

      I'm running the steady-state simulation first, then transferring those results to the transient simulation with flow freezing enabled.

      I've measured the atmospheric pressure, humidity and temperature in the area, and entered those values into the model.

      I've entered a table of temperature-dependent specific heat for the glass.

       

      Am I missing some parameter or material property that would improve the accuracy of the model?

       

      Thanks

        • Re: What's important?
          Jared Conway

          is radiation enabled?

           

          I'm running the steady-state simulation first, then transferring those results to the transient simulation with flow freezing enabled.

           

          >>why are you doing this?

           

          if after 2.5 minutes it gets down to 25dec C in a 25deg C environment, it is probably at steady state and steady state solution is your final answer

            • Re: What's important?
              Steve Calkins

              Radiation was not enabled.  My thinking was that it's a relatively cool process, so radiation wouldn't play as big a role.  I'll try it with radiation and see what the difference is.

               

              I'm using flow freezing to make the analysis run faster.  I've found on previous projects that trying to solve the transient without flow freezing took ridiculous amounts of time.  Previous projects and the SW kb suggest that it's a valid shortcut for my application. 

               

              The SS solution doesn't look at temperature at all, at least not the solid temperature.

               

              The transient solution looks at temperature.  In fact, the model has a solid representing the glass, and a solid representing the conveyor belt that carries the glass.  The belt model cools off, but the glass doesn't (well, it does, just not as much as in real life).

               

              The thought occurs to me that the conduction between the glass and the belt isn't being sufficiently considered.  This is a tough one to estimate.

              Would it be valid to just tweak this number until I get results that are closer?  I realize that would require assuming that all other parts of the simulation are accurate.

            • Re: What's important?
              Mark Keown

              Not sure I understand the configuration... but.

               

              Do you have the contact resistance between the between and the glass and belt correct?  I suspect in Flow the belt (large mass) is keeping the glass hot.  In reality the contact resistance between the belt and glass may be quite high.  Also if the cooling air is at a 'low' velocity definitely solve with radiation.  Lastly make sure that the TC wire is on the glass surface - for more than 20 diameters.  That is the TC bead is glued to the glass and TC wire glued or taped to the glass for a 20 mm or so.

                • Re: What's important?
                  Steve Calkins

                  Running again with radiation improved the results, but they're still different from reality by about 40C.

                   

                  I'm going to try again after adjusting the contact resistance between glass and belt. I think Mark is right, that in reality this resistance would be pretty high; glass is smooth, the belt is not.  But, not having nay actual data to enter for this, I'm having to make educated guesses.

                   

                  Thanks for your input. 

                    • Re: What's important?
                      Jared Conway

                      off by 40deg means in general that your setup doesn't match physical

                       

                      there are a dozen or so validation examples that you can run that show the results relative to hand calcs and physical that show how the software has been validated

                       

                      have you done some back of the envelope calcs just to make sure that with the conditions you have match what flow simulation is putting out?

                        • Re: What's important?
                          Steve Calkins

                          I have not done any hand calcs.  I have, however, done physical measurements on the system that was built from the SW model.

                          Trying to figure out which settings and conditions are important.  Adding radiation helped some, but not enough.  I've been fiddling with the contact resistance between the belt and the glass plate, and I am making some progress, but it's slow going: Adjust, run model, repeat...

                            • Re: What's important?
                              Jared Conway

                              i'd do some rough calcs just to see what is closer in the ball park

                               

                              going physical > sim is hard if you don't include all elements. the software has a lot of validation examples you can check if you are concerned that the software is not solving it properly. when you have more than 10% difference, something not being included is usually the cause.

                                • Re: What's important?
                                  Steve Calkins

                                  I'm sure the software is solving it correctly, in the sense that it's solving the equations with the information it's been given.  Garbage in, garbage out, as the saying goes.

                                   

                                  What I'm trying to figure out is what I haven't included in the model that I should have. 

                                  The software loses some value as a design tool if I can't get it to agree with actual measurements.

                                    • Re: What's important?
                                      Jared Conway

                                      we call what you're trying to do 3rd order accuracy where you adjust the software to the problem by eliminating assumptions. it takes awhile but if your goal is very accurate results, it is crucial. we're talking like 0.5degrees not 30, so I have a feeling there is something simple but large that you're missing. if you're interested in having someone look into it, we certainly could do that for you. jared@hawkridgesys.com